CLIMATE CHANGE - IT'S ALL ABOUT PROBABILITY

摘要

The assessment of present-day and future climate change is of crucial socio-economic and ecological importance but, at the same time, subject to a variety of uncertainty factors that are partly inherent to the climate system. This implies that a statement about the Earth's future climate is definitely a probabilistic one. From a scientific point of view, probabilistic statements require the knowledge of the probability density function (PDF) of the underlying process. In this paper, we expose what wealready know of the characteristics of such PDFs of climate change from coordinated climate modelling initiatives and what probabilistic statements can be derived from the quantification and evaluation of climate change in a regional and seasonal context. The first aspect addresses the changing occurrence of heat events with a relatively long return period in past climate. It turns out that particularly warm years, which only occurred once every 40 years in the past, will become typical events by the end of the 2fs' century. Thus, climate change can be perceived as a change in the probability of specific events. The second issue deals with the distinctness of past and future climates in the light of uncertainty. We show that this distinctness increases towards the end of our century and with the amount of greenhouse gas emissions. However, the overlapping probability of present-day and future PDFs still ranges between 10 and 30% for temperature and even beyond 90% for precipitation in some regions ofthe globe. The third problem is dedicated to so-called probabilistic climate predictions in the form of overshooting and undershooting probabilities given various thres'holds of climate change. While for temperature, the range of probable future changesis narrow and the sign is unambiguous, the uncertainty range of precipitation changes is often larger than the mean signal. Overall, probabilistic assessments in climate change research allow for the quantification of uncertainty and, hence, provide valuable information for decision processes.